If
you are interested in (1) IS implementation issues (2) IT and education (3)
Collaborative technology and (4) IT and culture, please contact A/P John Lim at
jlim@comp.nus.edu.sg
If
you are interested in the projects proposed by A/P Chee
Yam San, please look for him after
Talk to
new Research Student on
|
Time |
Speaker(s) |
Research Project Title |
|
|
A/P
Rudy Setiono |
Title: Support vector
machine for feature selection and classification Support vector machine (SVM) is a very useful
tool for pattern classification. It has
been shown to be very robust in many practical applications. In this project, we will be investigating
ways to select the relevant features involved in the discriminating hyperplane built by the support vector machine for a
given data set. The objective of feature selection is to find the smallest
subset of the features, and yet at the same time maintain the classification
accuracy. I
will describe SVM briefly, present my previous work on applying SVM for
credit risk evaluation, and discuss
potential research work in this area. |
|
|
Dr
Irene Woon |
Title: Socio-Technical
approach to IS Security A recent CSI/FBI survey (2002) shows that
security related losses have increased over the years, from US $100 million
in 1997 to US $455 million in 2002. This is despite the increased use of
security technologies such as firewalls and anti-virus software. Such figures
suggest that the narrow, technical-oriented solutions of most IS security
approaches are inadequate in meeting the security challenges of the Internet
age. We take a socio-technical
perspective view of IS security. Specifically, we see that human nature,
motivations, pyschology, organizational structures,
work cultures and norms have to be considered for IS security to be truly
effective. In this project, we seek to to
understand the social and technical factors that ensure information
confidentiality and integrity. We postulate that the following factors may
possibly have a role of play: Information availability, access control, responsbility and authority structures and information
ownerships. The candidate is expected to derive and test the relationship
between these and other factors so that practitioners may know the whole
spectrum of measures required to secure their data. |
|
|
Prof Krzysztof Apt |
Title: Two projects in
Constraint Programming Constraint programming is an alternative approach
to programming which relies on a combination of techniques that deal with
reasoning and computing. Research in this area focuses on various local
consistency notions (which approximate the global consistency notion),
algorithms designed to achieve them (called constraint propagation
algorithms), various types of constraint solvers (domain specific efficient
algorithms), and search methods. One of the most popular local consistency methods
deals with the removal of values from variable domains and is called arc
consistency. In the first part of the talk we shall explain its use and
suggest a research project dealing with it. In
the second part of the talk we shall discuss a research project concerned
with certain type of chess puzzles, such as http://www-gap.dcs.st-and.ac.uk/~history/Diagrams/Smullyan.gif
that could be solved by a combination of constraint programming and planning.
|
|
|
Dr |
Title: System Security and Multimedia In this talk, I will describe two problems: robust
signature and remote visualization. I will also briefly mention some activities in our research group. Robust signature can be used in the
authentication of multimedia data. For e.g. the photo in an identity
card. Given an object, say an image J,
we want to extract a signature S.
Suppose J1 is a slightly distorted version of J, we should able to
authenticate J1 with S
(robustness). On the
other hand, given S and
J, it is difficult for an
adversary to find another J2 that can be authenticated with S
(security). Our goal is to find such a scheme. In
remote visualization,
a server keeps a huge object, for example, data used in
image-based rendering. A client want to visualize this object. However, the client might not want to see
everything. So, we want to make use of
the client's intention to minimize the network and computation requirement. |
|
|
Dr
Terence Sim |
Title: Face modeling, rendering and recognition Human faces are important objects in image and video
understanding, and in computer graphics.
How do we model the face? How
do we acquire such models from photographs?
Can we render realistically from such models? How can we improve face recognition if we
know something about face shape? These
are challenging questions that we are investigating. The applications are many: better compressing/encoding techniques,
improved videoconferencing, robust biometrics, faster movie production,
etc. Our approach will use computer
graphics techniques together with computer vision and machine learning. Title: Improving digital photography Digital photography is popular these days, yet
the results are usually unsatisfactory.
We often encounter artifacts such as: red eye, uneven exposure, motion
blur, etc. While software like
Photoshop can help, the real solution lies in re-designing the camera,
combined with image processing algorithms.
We investigate such issues, using techniques from machine learning,
computer vision, and statistical signal processing. The goal is to create the next generation of
digital cameras: ones that produce perfect pictures automatically. Title: Music transcription Computers today can play music: that is, they can produce an audio signals from musical notation. Can it perform the inverse: going from audio signal to musical
score? The challenge is tremendous,
but the applications are many: from
music retrieval, to music education, to composition of new pieces. We will exploit music synthesis, signal
processing and machine learning techinques to solve
this and related problems. |
|
|
Dr
|
Title: Two Research
Problems in Streaming Media System In this talk I will give an overview of two
research problems in the area of multimedia systems. The first problem is how to improve the quality
of rendered environment when we stream 3D geometry models over best-effort
network in the presence of packet loss, delay jitter and bandwidth
constraint. The second problem concerns building a large
scale video sensors network for video surveillance. A system that intelligently and
efficiently manages the operations and interconnections of sensors, PCs and
monitors is required, as well as a formal framework for modeling and reasoning
about such a system. I
will highlight on-going, related research work in SoC,
and present possible future research directions. |
|
LUNCH
BREAK |
|
|
|
|
Dr
Antoine Vigneron |
Title: Computational
geometry projects Computational geometry is the study of algorithms
for solving geometric problems. The goal is usually to take advantage of the
geometric properties of the input (for instance, convexity) in order to
obtain a better running time. It finds applications in various areas of
computer science, such as spatial databases and graphics. In this talk, three research projects will be
presented. The first project is to design a provably good algorithm for a box
intersection problem. The second project is the implementation of a new
algorithm for a polygon decomposition problem. The third project is a
proximity problem for point sets in arbitrary dimension. |
|
|
|
|
|
|
Dr
Chan Chee Yong |
Title:
Topics in XML database research In this talk, I will describe three problems related to XML database research: content-based
data dissemination, XML query optimization, and indexing/clustering
of regular expressions. I will also highlight other interesting database
topics. |
|
|
Dr
Ken Sung |
Title:
Knowledge Discovery from Biological and Clinical Data There is a joint lab between SOC and I2R on
Computational Biology. The aim of the lab is to generating new
data-mining and algorithm techniques for solving problems in Computational
Biology. In this presentation, I will cover the areas of research interests
of the lab. |
Talk to
new Research Student on
|
Time |
Speaker(s) |
Research Project Title |
|
|
|
|
|
|
Dr
David Hsu |
Title: Geometric Computation
in Biology, Robotics, and Others With the rapid advances in technology,
computational methods are used increasingly to understand and interact with
the physical world. Many of these challenging problems require understanding
the geometric relationships among physical objects: *
How can the motion of digital movie actors be synthesized automatically? *
What can robots do to avoid running into people walking around? *
How do molecules change shapes over time to perform vital biological functions? In this talk, I will illustrate several
applications of geometric computation in computational biology, robotics, and
computer animation. |
|
|
A/P
|
Title: Intelligent
Document Text Processing In this talk, I will give an outline of our
document text processing work in CHIME.
In particular, there are two aspects of document text, one is in the
form of document images while the other in the form of electronic text. For the document images work, I will talk about a
problem of finding text in document images of widely different formats, some
mostly text, some mostly graphics while others may be somewhere in between.
The goal of this research is to find text without the prior knowledge of the
text/graphics composition. In the process of this research, we hope to build
up a document image database that contains a large collection of images with
ground truth similar to the well known For the electronic text, I will talk about a
joint research with I2R involving information extract from biology
literature. Due to the rapid advancement in life sciences, there is a wealth
of information in biological and medical literature. The work will focus on extracting important
information from abstracts of such literature, using techniques involving
machine learning, text processing and natural language understanding. Issues that will be examined include
anaphora resolution and named entity recognition. |
|
|
|
|
|
|
Dr
Shorey Rajeev |
Title: Investigations in
Multi-Hop Wireless Ad Hoc Networks |
Updated on
4/12/2003 11:08am
